PROJECT SUMMARY: PROJECT 4 The purpose of PROJECT 4 is to investigate computationally-informed precision treatments to improve two forms of state representation dysfunction in early psychosis: 1) State estimation processes at the perceptual input level, which we will target through auditory discrimination training; 2) State representation stability of auditory information, which we will target through auditory working memory training. Participants will be drawn from PROJECT 3, where they will have been assessed with behavioral and EEG-fMRI measures at baseline and after 6 months of usual care, so that their initial characteristics and clinical trajectory will be known. Participants will be stratified on an EEG index of state estimation processes (fronto-parietal theta power at DPX encoding), which we posit to be present in ~60% of subjects, and randomly assigned to one of the two training strategies. Our goal is not to perform a treatment efficacy study comparing these two interventions. Rather, we seek to use predictions derived from attractor network models to test the effects of neuroplasticity-based precision treatments targeting two distinct information processing pathologies in early psychosis, with the ultimate goal of improving state representation processes and cognition. In Aim 1, we will investigate parameter changes in the fit attractor network models in each subject group, fit to DPX and Bandit Task behavioral data immediately after training and 3 months later, and we will assess whether parameter changes reflect restorative or compensatory modifications. We will also test the hypothesis that state representation processes and cognitive performance show greater improvement in subjects who received training tailored to their state estimation parameter. In Aim 2, we will examine how specific parameter changes in attractor network models relate to neurophysiological changes in measures indexing activity timing, excitatory-inhibitory balance, and system noise, in order to identify which changes are the most predictive of improved cognition. Causal discovery analyses will be employed to identify causal relationships among computational parameters, behavioral data, neurophysiologic indices, treatment assignment, and one- year clinical trajectories.